AI Regulation Tracker  /  Legislation enacted

Taiwan passes its AI Basic Act: NSTC leads, high-risk duties left to sector regulators

Taiwan's Legislative Yuan cleared a national AI framework law on December 23, 2025. It sets seven governance principles and names a lead authority, but the hard duties for high-risk uses fall to each sector regulator, not the statute itself.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

Taiwan now has a national AI law. On December 23, 2025, the Legislative Yuan passed the Artificial Intelligence Basic Act (人工智慧基本法) on its third reading, giving the island a single statutory reference point for how the state and the private sector should build and use artificial intelligence. According to the Ministry of Digital Affairs (moda), the National Science and Technology Council (NSTC) began planning the measure in 2024 and drafted it after consulting industry, government, and academic stakeholders, with moda taking over legislative promotion in 2025.

The word that matters most for professionals is "basic." This is a framework statute, the kind Taiwan uses to set direction across many ministries rather than to impose specific commands. It carries no penalties of its own. That single fact reshapes how a lawyer, insurer, clinician, or compliance officer should read it.

What the law actually sets

The Act establishes seven governance principles: sustainable development and wellbeing, human autonomy, privacy and data governance, cybersecurity and safety, transparency and explainability, fairness and non-discrimination, and accountability. moda's summary describes the same principle set. These are the values every downstream rule and agency is expected to reflect. They are close cousins of the principles found in other national AI regimes, which makes the Taiwanese text familiar to anyone who has read the OECD principles or their regional adaptations.

The statute also names the NSTC as the central competent authority, the body responsible for coordinating national AI policy. moda is positioned as the ministry consulted on risk classification and is developing an AI risk framework to sit alongside the law.

Where the real duties will live

The Act signals that high-risk applications will face heavier obligations. The examples that recur in official and law-firm readings are self-driving vehicles, surgical robots, and screening systems used for admissions or hiring. For these, the direction of travel is warning labels, prior review, and third-party verification. The crucial detail is who sets those requirements. Under the Act, the sector regulator, the 中央目的事業主管機關, makes the high-risk designation in consultation with moda. In other words, the finance regulator decides for financial AI, the health authority for medical AI, and so on.

This is the operational heart of the story. The Basic Act does not tell a Taiwan hospital how to validate a diagnostic model or a bank how to disclose an AI chatbot. It tells the ministries that supervise them to produce those rules, consistent with the seven principles. Professionals who wait for the statute to spell out their duties will wait in vain. The duties will arrive as sector regulations.

What it does not do

The Act does not create fines, criminal liability, or a private right of action. It does not, by itself, ban any AI use or mandate any specific label with a defined penalty for non-compliance. moda's release describes the law as establishing the legislative foundation for governance frameworks to be developed afterward, and it does not set out enforcement mechanisms inside the statute. Reading direct obligations or numbers into the text would be a mistake. Where warning labels and prior review appear, they are commitments to be operationalized by sector rules, not self-executing commands.

The cross-border read

For a US professional, the contrast with the European Union is the useful frame. The EU AI Act imposes binding risk tiers and administrative fines directly in the instrument. Taiwan has chosen the opposite structure: a principles-first enabling statute that pushes hard duties down to sector regulators. That makes Taiwan a model to watch across the Asia-Pacific region, where several governments are weighing whether to legislate detailed obligations centrally or to steer through principles and let existing regulators fill in the specifics. It does not bind US firms, but any American company operating in Taiwan should expect its Taiwanese sector supervisor, not this law, to define what compliance looks like.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.